fix: wire --rebuild-index flag to bypass hash check, enabling ocr_redo regeneration

paperforge/commands/sync.py

@@ -62,6 +62,7 @@ def run(args: argparse.Namespace) -> int:
         "index_only": index_only,
         "prune": prune_flag,
         "prune_force": prune_force,
+        "rebuild_index": getattr(args, "rebuild_index", False),
     }
     try:
         sig = inspect.signature(svc.run)
@@ -82,7 +83,8 @@ def run(args: argparse.Namespace) -> int:
             except Exception:
                 pass
             try:
-                import subprocess, sys
+                import subprocess
+                import sys
                 subprocess.Popen(
                     [sys.executable, "-m", "paperforge", "embed", "build", "--resume"],
                     cwd=str(vault),
@@ -105,7 +107,8 @@ def run(args: argparse.Namespace) -> int:
         print(f"memory: deferred ({e})")
 
     try:
-        import subprocess, sys
+        import subprocess
+        import sys
         subprocess.Popen(
             [sys.executable, "-m", "paperforge", "embed", "build", "--resume"],
             cwd=str(vault),

paperforge/services/sync_service.py

@@ -200,7 +200,7 @@ class SyncService:
 
     def run(
         self, verbose: bool = False, json_output: bool = False, selection_only: bool = False, index_only: bool = False,
-        prune: bool = False, prune_force: bool = False,
+        prune: bool = False, prune_force: bool = False, rebuild_index: bool = False,
     ) -> PFResult:
         """Full sync orchestration. Returns PFResult contract.
 
@@ -229,9 +229,10 @@ class SyncService:
 
         # ── Phase 1: Select ──
         if not index_only:
-            from paperforge.worker.sync import run_selection_sync
             import time as _time
 
+            from paperforge.worker.sync import run_selection_sync
+
             _t0 = _time.time()
             try:
                 selection_result = run_selection_sync(self.vault, verbose=verbose, json_output=json_output)
@@ -288,7 +289,7 @@ class SyncService:
             flat_cleaned = self.clean_flat_notes(paths, json_output=json_output)
 
             if orphaned > 0 or flat_cleaned > 0:
-                index_count = asset_index.build_index(self.vault, verbose)
+            index_count = asset_index.build_index(self.vault, verbose, force_rebuild=rebuild_index)
 
             # ── Phase 4: Prune orphans ──
             prune_data = None

paperforge/worker/asset_index.py

@@ -524,17 +524,8 @@ def _compute_export_hash(paths: dict) -> str:
     return h.hexdigest()
 
 
-def build_index(vault: Path, verbose: bool = False) -> int:
+def build_index(vault: Path, verbose: bool = False, force_rebuild: bool = False) -> int:
     """Full rebuild of the canonical asset index for *vault*.
-
-    3-tier fast-path:
-    1. Computes a hash of BBT export JSON files.
-    2. Compares with ``export_hash`` stored in the existing index envelope.
-    3. If unchanged AND schema version matches, skips the entire rebuild loop.
-
-    Returns:
-        Number of items written to (or already present in) the index.
-    """
     from paperforge.config import load_vault_config
     from paperforge.worker._utils import pipeline_paths  # noqa: F811
     from paperforge.worker.base_views import ensure_base_views
@@ -559,7 +550,7 @@ def build_index(vault: Path, verbose: bool = False) -> int:
     zotero_dir = vault / cfg.get("system_dir", "System") / "Zotero"
 
     export_hash = _compute_export_hash(paths)
-    if isinstance(existing_data, dict) and existing_data.get("export_hash") == export_hash:
+    if not force_rebuild and isinstance(existing_data, dict) and existing_data.get("export_hash") == export_hash:
         if existing_data.get("schema_version") == CURRENT_SCHEMA_VERSION:
             count = len(existing_data.get("items", []))
             if verbose: